Autoregressive modeling application for vascular zones detection in the contrast echographic images

Classification of Contrast Ultrasound Images using Autoregressive Model Coupled to Gaussian Mixture Model

“…First, the ATR and FISHER values will be better, and secondly, this should lead to the agent density quantification and thus helps the echographic technician in the diagnostic. This technique does not need to use classification methods as the precedent methods [6]. However, the great advantage of this stage relies on the fact that the threshold value, which has to be determined empirically after many tests, is now determined automatically by the Gaussian model during the classification step.…”

“…In this technique, the AR-based covariance has been selected to calculate the prediction error since it has been shown to be the best method among the other AR-based approaches [6].…”

“…However, this weak nonlinearity may be considered as linear [7] in comparison with the agentultrasound interaction which is characterized by a strong nonlinear behavior [8]. We proposed in a previous work [6] to apply the autoregressive (AR) model [9] on the backscattered signal to separate the tissue part and the agent part. Our approach is mainly based on the autoregressive prediction error to enhance the contrast between the two parts.…”

“…Our approach is mainly based on the autoregressive prediction error to enhance the contrast between the two parts. We propose to implement a new approach based on the AR-covariance method [6] coupled to the Gaussian mixture model (GMM) that shows better performance than the harmonic mode [4] which is implemented in most echographic systems. The obtained results show that the prediction error is relatively large for the signal issued from contrast agents when compared to that of the tissues.…”